Newsgroups: comp.mail.misc
Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!uupsi!schoff
From: schoff@uu.psi.com (Martin Schoffstall)
Subject: Re: mcimail.com: Where'd it go?
Message-ID: <1991May16.034406.6831@uu.psi.com>
Organization: Performance Systems International, Inc.
References: <17896@venera.isi.edu> <kc_fpwj0BwwOF=Rpov@transarc.com>
Date: Thu, 16 May 91 03:44:06 GMT

Craig,

I've been a little remiss in keeping up with USENET this last two
weeks so this is a late response.....

As you know the DNS is pretty asynchronous, and convergence of
 servers takes time in general.....

This is also true at the root level, the few people left at the
NIC are working very hard, and considering the morale they
must have at SRI due to the transfer of the NIC to a newly
created company who was the low bidder, they are doing great,
but they make mistakes... mistakes in data exported to
the roots.

The roots (c.nyser.net is what we run) are slaves of the data
from the NIC, half of the roots are controlled by the NIC
directly, the other half are not (ns.nasa.gov, c.nyser.net)
aren't.  By agreement, we only load data from the NIC,
we don't patch on the fly.  This is all background so far...

Regardless of the control issue the root's have different data
from time to time (hence the different SOA's), eventually
they converge....

between data sets we've seen the following from time to time

Monday AM  - MYDOMAIN.com available in SOA N

Wednesday AM  - MYDOMAIN.com removed in SOA N+1

(people complain/scream/beg)

Friday AM - MYDOMAIN available in SOA N+2

Friday Afternoon - MYDOMAIN still not usable due to
	(a) roots not in sync
	(b) caching of NXDOMAIN in some number of hosts

Marty

In article <kc_fpwj0BwwOF=Rpov@transarc.com> Craig_Everhart@TRANSARC.COM writes:
>There are intermittent problems resolving name ``mcimail.com''.  There
>are seven name servers for COM.  When I just queried each of them, two
>listed ``mcimail.com'' with no problems, two gave authoritative NXDOMAIN
>responses, and three were unavailable.  The correct servers
>(ns.nic.ddn.mil aos.brl.mil) had a later serial number for the COM zone
>than did the erroneous servers (c.nyser.net, ns.nasa.gov), so the
>problem will correct itself in time.  However, it could be appropriate
>for the managers of the erroneous servers to trigger a zone update if
>they can.
>
>		Thanks,
>		Craig Everhart
>
>ns.nic.ddn.mil: worked, soa(com) 910509
>aos.brl.mil: worked, soa(com) 910509
>a.isi.edu: timeout
>c.nyser.net: failed, serial(com) 910506
>terp.umd.edu: timeout
>ns.nasa.gov: failed, serial(com) 910506
>gunter-adam.af.mil: timeout


